And I'm sure that services will totally not get worse when you remove the last human in the loop from the money extraction machine that is the health insurance industry.
And I'm sure that services will totally not get worse when you remove the last human in the loop from the money extraction machine that is the health insurance industry.
Think of it this way: is it really an efficient allocation of economic resources to have so many educated people working in healthcare bureaucracy? As the essay points out, the number of admin...
Think of it this way: is it really an efficient allocation of economic resources to have so many educated people working in healthcare bureaucracy? As the essay points out, the number of admin have grown at a 10x rate that of physicians.
A good bureaucracy has lots of people to insure things actually get done. A bad bureaucracy has 1 person taking the blame for an AI fucking up 10,000 faster than a human. Firing the people in the...
A good bureaucracy has lots of people to insure things actually get done. A bad bureaucracy has 1 person taking the blame for an AI fucking up 10,000 faster than a human. Firing the people in the loop just results in a worse bureaucracy.
Have you ever been able to resolve an edge-case problem quickly by clicking through 10 layers of phone prompts? No, you get the problem solved by talking to an actual human.
Everything service I use that has deployed AI agents to "help customers" uses it as an excuse to make the service even shittier. Because AI will handle the FAQ fine. But if your problem isn't on the FAQ you're SOL.
The proper way to improve efficiency is to murder the health insurance industry.
My doc has said he could double his patient count if he only took Medicare or cash.
I've heard a good chunk of hospital admin positions exist solely to deal with the complicated minefield that is the US healthcare insurance industry. They have different requirements and policies,...
I've heard a good chunk of hospital admin positions exist solely to deal with the complicated minefield that is the US healthcare insurance industry. They have different requirements and policies, so even simple and routine bills take up time because they have to double-check that the paperwork fits that particular insurer's requirements.
That said, it feels like AI might be able to help in that particular area a little bit? Not by replacing humans entirely, but handling the lower level stuff (e.g. filling out the routine paperwork and ensuring codes match the respective companies' standards). That would leave humans more time to handle negotiating and arguing for the more complex cases.
At the very least, that unfortunately seems like a more actually feasible course of action to maybe reduce costs and reallocate funding than overhauling or killing the health insurance industry.
Besides that though... From what I've gathered, even outside insurance, there's a lot of bloat in hospital bureaucracy right now. Most of my knowledge is purely anecdotal, but at the peak of the pandemic I remember many, many, many nurses venting about the admin bloat at their hospitals and clinics on reddit. Particularly at the corporate level, but even admins in their buildings caused frustration because they couldn't/wouldn't advocate for their staff to corporate due to a mix of selfish priorities and just not understanding what the floor is actually like.
It actually hinders the medical staff's ability to do their jobs because people with no experience with patients were in charge of everything. Putting out impossible schedules, keeping facilities understaffed, doing the bare minimum in ordering and distributing equipment, organizing meetings over trivial things that genuinely wasted time but looked good for the admins, different departments lobbing responsibilities back and forth like tennis...
Honestly the unfortunate thing is that the ones causing the most problems seem to be those who are harder to replace by AI. There's definitely something really messed up with the bureaucratic side of American hospitals beyond dealing with insurance.
What you're describing though is more like saying "Boy, it really sucks having to chop off all these hands with axes. Let's use a woodchipper instead, it'll be much more efficient!" Nobody...
What you're describing though is more like saying "Boy, it really sucks having to chop off all these hands with axes. Let's use a woodchipper instead, it'll be much more efficient!" Nobody considers that maybe we should stop chopping off hands. Automating with AI will not make work for doctors and nurses easier. The experience will get worse for patients. And the only people really benefiting are the managers whom are abstracted from the consequences of both, and get to have heavy bonuses for cutting headcount.
Another anecdote, but everything I've seen regarding bad bureaucracy pretty much everywhere is a misalignment of incentives, and they tend to be worse when profit incentives are paramount. It's almost like healthcare in particular should not be permitted to be for-profit, because providing good service is orthogonal to increasing profits.
Cory Doctorow has highlighted an important facet that is talked about in automation: The Centaur vs the Reverse Centaur. He used the example of AI X-Ray technicians. In the Centaur model, the same number of X-Ray technicians remain, and the AI is used to supplement their expertise and provide better service. In the Reverse Centaur model, a single X-Ray tech now just audits the machines expertise, and serves primarily as a blame sink for when the AI gets it wrong (incidentally the AI model will stop improving if there isn't a solid improving feedback loop).
Being a Reverse Centaur sucks. Because you get fired if you fuck up because you're overworked and don't notice the tiniest mistake the AI makes. And you have no job autonomy. Amazon warehouse workers are much like this.
It is an important distinction, and it is one that is at the heart of why the Luddites were right: Machines should be used as tools for artisans. They should not be a way of replacing artisans with wage slaves.
That has been, broadly speaking, the general truism with the rise of industrialization. The goal is rarely to provide a better good/service; It's to churn out more at the same or lower quality...
That has been, broadly speaking, the general truism with the rise of industrialization. The goal is rarely to provide a better good/service; It's to churn out more at the same or lower quality with lower costs but same prices.
Even from the beginning, the deployment of machines was used by the capitalists to replace skilled labor with unskilled labor wherever possible.
And the sales pitch for AI to the execs is exactly that: Fire half your staff, you give us half and you keep half.
Reducing costs (and prices) is obviously an important goal, but it's also the case that many products have vastly improved. Computers, electronics and cars most obviously.
Reducing costs (and prices) is obviously an important goal, but it's also the case that many products have vastly improved. Computers, electronics and cars most obviously.
Never tried it, but I've heard that asking for the cash price sometimes works? We do pay cash for dental and vision, because it wasn't clear that the insurance was worth it.
Never tried it, but I've heard that asking for the cash price sometimes works?
We do pay cash for dental and vision, because it wasn't clear that the insurance was worth it.
I'm not going to defend the administrative bloat of many systems. But there is little distinction between a system with excess capacity as a healthy redundancy; and systems that are bogged down...
I'm not going to defend the administrative bloat of many systems. But there is little distinction between a system with excess capacity as a healthy redundancy; and systems that are bogged down with excessive costs, meaningless middle-men, predatory contracts and being extorted by cronies. The former represents a system that provides stable employment and operational flexibility. The latter is sadly the reality that we are dealing with and the few functional actors in the system are over-extended to mitigate all the waste around them.
And the way I see it, AI is an attempt to replace those functional actors with an automated system that are the embodiment of excessive costs, meaningless middle-men, predatory contracts and cronies. All justified by the idea of "efficiency" and "cost effectiveness". Promises that are completely false.
I was interviewing someone that runs an insurance brokerage and he said that scaling with people is a linear cost while AI is flat. And it hit me on just how well obscured the costing was for average users. As if you can just pay openAI a fixed monthly cost and you could just spin up as many LLM instances as you want.
AI now is still in the first days of Uber and AirBnB. The company is going to eat a ton of the cost and plan to recoup it with interest once they capture the market and establish user dependency. True cost of LLM's often scale with the context size because the system reprocesses the entire instance with every time its invoked. Simpler analytical models can be linear but its often exponential. Advanced Multi-model and Agent systems multiply those costs.
Probably not, but finding another job isn't easy. What businesses are expanding? What do they need people to do? A lot of the job creators want to use AI instead.
Probably not, but finding another job isn't easy. What businesses are expanding? What do they need people to do? A lot of the job creators want to use AI instead.
It's the age old problem with automation. It's supposed to make people's lives better, and for the most part is has; but the damage to the peoples' whose jobs it replaces is real and isn't easy to...
It's the age old problem with automation. It's supposed to make people's lives better, and for the most part is has; but the damage to the peoples' whose jobs it replaces is real and isn't easy to recover from for a long time.
Historically automation only affected "low skill" labor so it might have been possible for people to "trade up to better jobs" so to speak. We've never really been in a position where automation was threatening so many "high skill" jobs, so its hard to say what the future looks like. Ideally the end goal is a society where people don't have to work (or at least don't have to work on anywhere near the level they do today,) but I just don't see how that ever becomes possible unless there is a drastic shift in how the profits from automating all these jobs away gets shared. Right now the only people "winning" by jobs getting automated away are the shareholders.
The future I want to see is one where people perform a few "tours of duty" through their lives, with the rest of their time spent not working. So between 21 and 65 years old you'll be expected to...
The future I want to see is one where people perform a few "tours of duty" through their lives, with the rest of their time spent not working. So between 21 and 65 years old you'll be expected to do 15 years full time. You might spend one "career" doing 5 years helping the sick, 5 years building infrastructure, and 5 years maintaining infrastructure with optionally some time off in between each stint.
People will of course find good things to do with their time off of work. I haven't worked much for money since August 2024 and ended up volunteering some time each day to help with a non-profit in my city.
I think you run out of, among others, mental health and medical professionals pretty fast that way. There are any number of pretty specialized fields like that and it just wouldn't make sense to...
I think you run out of, among others, mental health and medical professionals pretty fast that way. There are any number of pretty specialized fields like that and it just wouldn't make sense to me to have those folks work 5-15 years as a social worker. And I wouldn't want to work like that myself.
I love my job, and it's the field I want to work in. Unless we're post singularity to the extent that no one has any problems anymore I think we'd always need those fields.
I manage the automation team in a Canadian public health organization. The opportunities are endless. Ironically given the mention of AI scribes - we are now experiencing the “chart bloat” that...
I manage the automation team in a Canadian public health organization. The opportunities are endless.
Ironically given the mention of AI scribes - we are now experiencing the “chart bloat” that comes from these. The previous transcribers would do a good job of summarizing the contents of the dictations. We’re now looking at LLM based medical record summarization to defeat the LLM based bloat.
But there are tons of use cases, I spend much of my week on the phone to people discussing their pain points. In public healthcare our problem isn’t so much all the needless billing related stuff, it’s because we’re so strapped for discretionary funds, it’s easier to hire a body to work the shitty system then it is to obtain and invest capital to fix the shitty system. Lots and lots of technical debt leading to overhead and unnecessary roles.
I guess I'd be curious to hear what kind of things you and your team think would be actually beneficial to automate, vs what execs think can be fixed by "just using AI". (I'm in the US, so know...
AMA I guess?
I guess I'd be curious to hear what kind of things you and your team think would be actually beneficial to automate, vs what execs think can be fixed by "just using AI".
(I'm in the US, so know very little about Canadian healthcare beyond anecdotes from Candadian friends)
Well I can tell you that we're saving (as defined as, automated transactions per period * avg manual handling time of each transaction) tens of thousands of hours per year and we're doing it...
Well I can tell you that we're saving (as defined as, automated transactions per period * avg manual handling time of each transaction) tens of thousands of hours per year and we're doing it without using AI. We're doing tons of "classical" process automation, in other words deterministic software workflows using a mixture of API automation and GUI automation.
The AI vendors (and hell, even the legacy automation companies) are trying to sell us on the concept that pure agentic/AI driven automation is the future, but having watched the state of the art computer-using agents attempt to reason themselves out of a paper bag and only occasionally succeed, and do so extremely slowly, I will take automations written by my developers that are able to process thousands of transactions per day, and do it deterministically/reliably every time. It's not sexy, but in large, SOP-driven enterprises like ours, volume wins the day.
That being said, I think there is a place for AI in automation, and natural language processing is it. Some example use cases we've been exploring, other than the record summarization I mentioned:
Protocolling requisitions (i.e. automating the work of technicians who read a req and match up what the clinician has requested vs exam types available, check for any contraindications and so on)
Scheduling low-priority procedures via outbound voice calls where digital self-scheduling isn't working out for some reason
Handling tier 1 IT service desk emails and chats (not a healthcare-specific use case, but we have this problem just like everyone else)
The magic sauce in all of these is the ability to parse and generate language.
It's not about replacing jobs (well, usually. The IT one is about reducing subcontractor spend), it's about moving patients through the system faster and more efficiently, and ultimately providing better care.
My prediction is that AI will lead to the loss of hundreds of thousands of healthcare jobs, but few of these will be among physicians, allied health professionals (AHPs), and nurses, at least in the next decade.
...
Wander into the billing department of most large health systems, and you’ll witness hundreds of people optimizing documentation, drafting prior auth requests, and otherwise pushing reams of paper, often into fax machines. Amble over to the quality department, where you’ll see scores of workers reviewing charts for data to populate quality measurement spreadsheets, pay-for-performance schemes, and national registries in areas like interventional cardiology and transplantation. Head over to the population health department, whose personnel are struggling to remind patients that they’re overdue for a mammogram or a colonoscopy. The call center… well, you get the picture.
...
While administrative and operational jobs have expanded most rapidly, they’re also the most susceptible to AI replacement. The cuts have already begun. At UCSF Health, a decade ago, we started hiring human scribes to assist our busiest ambulatory clinicians in documenting their patient visits. At the program’s peak, about 100 of our busiest clinicians had a scribe – often a pre-medical student on a gap year. Today at UCSF, thousands of our doctors are now using scribes – but it’s no longer a 22-year-old pre-med supplied by a scribe staffing company; it’s an AI scribe supplied by a company named Abridge. Our clinicians are quite happy with their AI scribes (though some tell me they miss the opportunity to mentor the human scribes), and our expenses are a fraction of what they once were.
...
In contrast to the major cuts in administrative staff, I predict that very few doctors, nurses, or AHPs will lose their jobs, at least over the next decade. For nurses in the hospital, much of their work involves administering medications, assisting patients with toileting and mobility, inserting catheters and IVs, conducting complex assessments, counseling patients and their families, and responding to emergencies. For the foreseeable future, it’s more likely that AI and robots will assist nurses in performing their jobs more effectively – enabling them to practice closer to the top of their license – than replace those jobs entirely.
For physicians, even fields seen as highly vulnerable to job replacement by AI (particularly radiology and pathology) have proven exceptionally resilient. If you’d asked me 15 years ago which would come first: our radiologists would be unemployed, or I’d sit in the back seat of a driverless car without giving it a second thought, I would have assumed that the radiologists would be toast. Yet in San Francisco, we can’t hire radiologists fast enough, while Waymos are literally everywhere. (In my book, I describe why radiologists have been surprisingly shielded from job replacement.)
And I'm sure that services will totally not get worse when you remove the last human in the loop from the money extraction machine that is the health insurance industry.
Think of it this way: is it really an efficient allocation of economic resources to have so many educated people working in healthcare bureaucracy? As the essay points out, the number of admin have grown at a 10x rate that of physicians.
A good bureaucracy has lots of people to insure things actually get done. A bad bureaucracy has 1 person taking the blame for an AI fucking up 10,000 faster than a human. Firing the people in the loop just results in a worse bureaucracy.
Have you ever been able to resolve an edge-case problem quickly by clicking through 10 layers of phone prompts? No, you get the problem solved by talking to an actual human.
Everything service I use that has deployed AI agents to "help customers" uses it as an excuse to make the service even shittier. Because AI will handle the FAQ fine. But if your problem isn't on the FAQ you're SOL.
The proper way to improve efficiency is to murder the health insurance industry.
My doc has said he could double his patient count if he only took Medicare or cash.
I've heard a good chunk of hospital admin positions exist solely to deal with the complicated minefield that is the US healthcare insurance industry. They have different requirements and policies, so even simple and routine bills take up time because they have to double-check that the paperwork fits that particular insurer's requirements.
That said, it feels like AI might be able to help in that particular area a little bit? Not by replacing humans entirely, but handling the lower level stuff (e.g. filling out the routine paperwork and ensuring codes match the respective companies' standards). That would leave humans more time to handle negotiating and arguing for the more complex cases.
At the very least, that unfortunately seems like a more actually feasible course of action to maybe reduce costs and reallocate funding than overhauling or killing the health insurance industry.
Besides that though... From what I've gathered, even outside insurance, there's a lot of bloat in hospital bureaucracy right now. Most of my knowledge is purely anecdotal, but at the peak of the pandemic I remember many, many, many nurses venting about the admin bloat at their hospitals and clinics on reddit. Particularly at the corporate level, but even admins in their buildings caused frustration because they couldn't/wouldn't advocate for their staff to corporate due to a mix of selfish priorities and just not understanding what the floor is actually like.
It actually hinders the medical staff's ability to do their jobs because people with no experience with patients were in charge of everything. Putting out impossible schedules, keeping facilities understaffed, doing the bare minimum in ordering and distributing equipment, organizing meetings over trivial things that genuinely wasted time but looked good for the admins, different departments lobbing responsibilities back and forth like tennis...
Honestly the unfortunate thing is that the ones causing the most problems seem to be those who are harder to replace by AI. There's definitely something really messed up with the bureaucratic side of American hospitals beyond dealing with insurance.
What you're describing though is more like saying "Boy, it really sucks having to chop off all these hands with axes. Let's use a woodchipper instead, it'll be much more efficient!" Nobody considers that maybe we should stop chopping off hands. Automating with AI will not make work for doctors and nurses easier. The experience will get worse for patients. And the only people really benefiting are the managers whom are abstracted from the consequences of both, and get to have heavy bonuses for cutting headcount.
Another anecdote, but everything I've seen regarding bad bureaucracy pretty much everywhere is a misalignment of incentives, and they tend to be worse when profit incentives are paramount. It's almost like healthcare in particular should not be permitted to be for-profit, because providing good service is orthogonal to increasing profits.
Cory Doctorow has highlighted an important facet that is talked about in automation: The Centaur vs the Reverse Centaur. He used the example of AI X-Ray technicians. In the Centaur model, the same number of X-Ray technicians remain, and the AI is used to supplement their expertise and provide better service. In the Reverse Centaur model, a single X-Ray tech now just audits the machines expertise, and serves primarily as a blame sink for when the AI gets it wrong (incidentally the AI model will stop improving if there isn't a solid improving feedback loop).
Being a Reverse Centaur sucks. Because you get fired if you fuck up because you're overworked and don't notice the tiniest mistake the AI makes. And you have no job autonomy. Amazon warehouse workers are much like this.
It is an important distinction, and it is one that is at the heart of why the Luddites were right: Machines should be used as tools for artisans. They should not be a way of replacing artisans with wage slaves.
This is assuming that new technologies will be implemented badly. Often they are, but I think that's better judged on a case-by-case basis.
That has been, broadly speaking, the general truism with the rise of industrialization. The goal is rarely to provide a better good/service; It's to churn out more at the same or lower quality with lower costs but same prices.
Even from the beginning, the deployment of machines was used by the capitalists to replace skilled labor with unskilled labor wherever possible.
And the sales pitch for AI to the execs is exactly that: Fire half your staff, you give us half and you keep half.
Reducing costs (and prices) is obviously an important goal, but it's also the case that many products have vastly improved. Computers, electronics and cars most obviously.
As a thankfully healthy young person I really wish hospitals just took cash at reasonable market rates.
Never tried it, but I've heard that asking for the cash price sometimes works?
We do pay cash for dental and vision, because it wasn't clear that the insurance was worth it.
I'm not going to defend the administrative bloat of many systems. But there is little distinction between a system with excess capacity as a healthy redundancy; and systems that are bogged down with excessive costs, meaningless middle-men, predatory contracts and being extorted by cronies. The former represents a system that provides stable employment and operational flexibility. The latter is sadly the reality that we are dealing with and the few functional actors in the system are over-extended to mitigate all the waste around them.
And the way I see it, AI is an attempt to replace those functional actors with an automated system that are the embodiment of excessive costs, meaningless middle-men, predatory contracts and cronies. All justified by the idea of "efficiency" and "cost effectiveness". Promises that are completely false.
I was interviewing someone that runs an insurance brokerage and he said that scaling with people is a linear cost while AI is flat. And it hit me on just how well obscured the costing was for average users. As if you can just pay openAI a fixed monthly cost and you could just spin up as many LLM instances as you want.
AI now is still in the first days of Uber and AirBnB. The company is going to eat a ton of the cost and plan to recoup it with interest once they capture the market and establish user dependency. True cost of LLM's often scale with the context size because the system reprocesses the entire instance with every time its invoked. Simpler analytical models can be linear but its often exponential. Advanced Multi-model and Agent systems multiply those costs.
Probably not, but finding another job isn't easy. What businesses are expanding? What do they need people to do? A lot of the job creators want to use AI instead.
It's the age old problem with automation. It's supposed to make people's lives better, and for the most part is has; but the damage to the peoples' whose jobs it replaces is real and isn't easy to recover from for a long time.
Historically automation only affected "low skill" labor so it might have been possible for people to "trade up to better jobs" so to speak. We've never really been in a position where automation was threatening so many "high skill" jobs, so its hard to say what the future looks like. Ideally the end goal is a society where people don't have to work (or at least don't have to work on anywhere near the level they do today,) but I just don't see how that ever becomes possible unless there is a drastic shift in how the profits from automating all these jobs away gets shared. Right now the only people "winning" by jobs getting automated away are the shareholders.
The future I want to see is one where people perform a few "tours of duty" through their lives, with the rest of their time spent not working. So between 21 and 65 years old you'll be expected to do 15 years full time. You might spend one "career" doing 5 years helping the sick, 5 years building infrastructure, and 5 years maintaining infrastructure with optionally some time off in between each stint.
People will of course find good things to do with their time off of work. I haven't worked much for money since August 2024 and ended up volunteering some time each day to help with a non-profit in my city.
I think you run out of, among others, mental health and medical professionals pretty fast that way. There are any number of pretty specialized fields like that and it just wouldn't make sense to me to have those folks work 5-15 years as a social worker. And I wouldn't want to work like that myself.
I love my job, and it's the field I want to work in. Unless we're post singularity to the extent that no one has any problems anymore I think we'd always need those fields.
I manage the automation team in a Canadian public health organization. The opportunities are endless.
Ironically given the mention of AI scribes - we are now experiencing the “chart bloat” that comes from these. The previous transcribers would do a good job of summarizing the contents of the dictations. We’re now looking at LLM based medical record summarization to defeat the LLM based bloat.
But there are tons of use cases, I spend much of my week on the phone to people discussing their pain points. In public healthcare our problem isn’t so much all the needless billing related stuff, it’s because we’re so strapped for discretionary funds, it’s easier to hire a body to work the shitty system then it is to obtain and invest capital to fix the shitty system. Lots and lots of technical debt leading to overhead and unnecessary roles.
AMA I guess?
I guess I'd be curious to hear what kind of things you and your team think would be actually beneficial to automate, vs what execs think can be fixed by "just using AI".
(I'm in the US, so know very little about Canadian healthcare beyond anecdotes from Candadian friends)
Well I can tell you that we're saving (as defined as, automated transactions per period * avg manual handling time of each transaction) tens of thousands of hours per year and we're doing it without using AI. We're doing tons of "classical" process automation, in other words deterministic software workflows using a mixture of API automation and GUI automation.
The AI vendors (and hell, even the legacy automation companies) are trying to sell us on the concept that pure agentic/AI driven automation is the future, but having watched the state of the art computer-using agents attempt to reason themselves out of a paper bag and only occasionally succeed, and do so extremely slowly, I will take automations written by my developers that are able to process thousands of transactions per day, and do it deterministically/reliably every time. It's not sexy, but in large, SOP-driven enterprises like ours, volume wins the day.
That being said, I think there is a place for AI in automation, and natural language processing is it. Some example use cases we've been exploring, other than the record summarization I mentioned:
The magic sauce in all of these is the ability to parse and generate language.
It's not about replacing jobs (well, usually. The IT one is about reducing subcontractor spend), it's about moving patients through the system faster and more efficiently, and ultimately providing better care.
From the article:
...
...
...